Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
156 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (2306.06963v3)

Published 12 Jun 2023 in cs.CV

Abstract: The imbalanced distribution of long-tailed data presents a considerable challenge for deep learning models, as it causes them to prioritize the accurate classification of head classes but largely disregard tail classes. The biased decision boundary caused by inadequate semantic information in tail classes is one of the key factors contributing to their low recognition accuracy. To rectify this issue, we propose to augment tail classes by grafting the diverse semantic information from head classes, referred to as head-to-tail fusion (H2T). We replace a portion of feature maps from tail classes with those belonging to head classes. These fused features substantially enhance the diversity of tail classes. Both theoretical analysis and practical experimentation demonstrate that H2T can contribute to a more optimized solution for the decision boundary. We seamlessly integrate H2T in the classifier adjustment stage, making it a plug-and-play module. Its simplicity and ease of implementation allow for smooth integration with existing long-tailed recognition methods, facilitating a further performance boost. Extensive experiments on various long-tailed benchmarks demonstrate the effectiveness of the proposed H2T. The source code is available at https://github.com/Keke921/H2T.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (64)
  1. Deep Over-sampling Framework for Classifying Imbalanced Data. In ECML/PKDD, 770–785.
  2. ACE: Ally Complementary Experts for Solving Long-Tailed Recognition in One-Shot. In ICCV, 112–121.
  3. Learning imbalanced datasets with label-distribution-aware margin loss. In NeurIPS, 1567–1578.
  4. SMOTE: Synthetic Minority Over-sampling Technique. Journal of artificial intelligence research, 16: 321–357.
  5. Facial Structure Guided GAN for Identity-Preserved Face Image De-Occlusion. In ICMR, 46–54.
  6. Feature Space Augmentation for Long-Tailed Data. In ECCV, volume 12374, 694–710.
  7. AutoAugment: Learning Augmentation Strategies From Data. In CVPR, 113–123.
  8. Randaugment: Practical automated data augmentation with a reduced search space. In CVPRW, 3008–3017.
  9. ResLT: Residual Learning for Long-Tailed Recognition. IEEE TPAMI, 45(3): 3695–3706.
  10. Parametric contrastive learning. In CVPR, 715–724.
  11. Class-balanced loss based on effective number of samples. In CVPR, 9268–9277.
  12. ArcFace: Additive Angular Margin Loss for Deep Face Recognition. In CVPR, 4690–4699.
  13. Deep residual learning for image recognition. In CVPR, 770–778.
  14. Population Based Augmentation: Efficient Learning of Augmentation Policy Schedules. In ICML, volume 97, 2731–2741.
  15. Disentangling Label Distribution for Long-Tailed Visual Recognition. In CVPR, 6626–6636.
  16. Joint Semantic Preserving Sparse Hashing for Cross-Modal Retrieval. IEEE TCSVT, 1–15.
  17. Triplet Fusion Network Hashing for Unpaired Cross-Modal Retrieval. In ICMR, 141–149.
  18. Learning Deep Representation for Imbalanced Classification. In CVPR.
  19. Deep Imbalanced Learning for Face Recognition and Attribute Prediction. IEEE TPAMI, 42(11): 2781–2794.
  20. An Optimal Transport View of Class-Imbalanced Visual Recognition. International Journal of Computer Vision, 1–19.
  21. Long-Tailed Visual Recognition via Self-Heterogeneous Integration With Knowledge Excavation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 23695–23704.
  22. Decoupling representation and classifier for long-tailed recognition. In ICLR.
  23. M2m: Imbalanced classification via major-to-minor translation. In CVPR, 13896–13905.
  24. Learning multiple layers of features from tiny images. Tech Report.
  25. Compact Neural Network via Stacking Hybrid Units. IEEE TPAMI, 46(1): 103–116.
  26. Trustworthy Long-Tailed Classification. In CVPR, 6970–6979.
  27. Nested Collaborative Learning for Long-Tailed Visual Recognition. In CVPR, 6949–6958.
  28. Key Point Sensitive Loss for Long-tailed Visual Recognition. IEEE TPAMI, in Press: 1–14.
  29. Long-Tailed Visual Recognition via Gaussian Clouded Logit Adjustment. In CVPR, 6929–6938.
  30. Adjusting Logit in Gaussian Form for Long-Tailed Visual Recognition. arXiv preprint arXiv:2305.10648.
  31. MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognition. In CVPR, 5212–5221.
  32. Fast AutoAugment. In NeurIPS, 6662–6672.
  33. Focal Loss for Dense Object Detection. IEEE TPAMI, 42(2): 318–327.
  34. Memory-Based Jitter: Improving Visual Recognition on Long-Tailed Data with Diversity in Memory. In AAAI, 1720–1728. AAAI Press.
  35. Deep Representation Learning on Long-Tailed Data: A Learnable Embedding Augmentation Perspective. In CVPR.
  36. MTFH: A Matrix Tri-Factorization Hashing Framework for Efficient Cross-Modal Retrieval. IEEE TPAMI, 43(3): 964–981.
  37. Large-Scale Long-Tailed Recognition in an Open World. In CVPR, 2537–2546.
  38. Image Segmentation Using Deep Learning: A Survey. IEEE TPAMI, 44(7): 3523–3542.
  39. The Majority Can Help The Minority: Context-rich Minority Oversampling for Long-tailed Classification. In CVPR, 6887–6896.
  40. Influence-Balanced Loss for Imbalanced Visual Classification. In ICCV, 735–744.
  41. Reed, W. J. 2001. The Pareto, Zipf and other power laws. Economics letters, 74(1): 15–19.
  42. Balanced Meta-Softmax for Long-Tailed Visual Recognition. In NeurIPS, volume 33, 4175–4186.
  43. Learning to Reweight Examples for Robust Deep Learning. In ICML, volume 80, 4331–4340.
  44. ImageNet Large Scale Visual Recognition Challenge. IJCV, 115(3): 211–252.
  45. Going deeper with convolutions. In CVPR, 1–9.
  46. The INaturalist Species Classification and Detection Dataset. In CVPR, 8769–8778.
  47. Manifold mixup: Better representations by interpolating hidden states. In ICML, 6438–6447.
  48. NormFace: L22{}_{\mbox{2}}start_FLOATSUBSCRIPT 2 end_FLOATSUBSCRIPT Hypersphere Embedding for Face Verification. In ACM MM, 1041–1049.
  49. Cosface: Large margin cosine loss for deep face recognition. In CVPR, 5265–5274.
  50. RSG: A Simple but Effective Module for Learning Imbalanced Datasets. In CVPR, 3784–3793.
  51. Long-tailed Recognition by Routing Diverse Distribution-Aware Experts. In ICLR.
  52. Learning From Multiple Experts: Self-paced Knowledge Distillation for Long-Tailed Classification. In ECCV, 247–263.
  53. Does head label help for long-tailed multi-label text classification. AAAI, 35(16): 14103–14111.
  54. A survey on long-tailed visual recognition. IJCV, 130(7): 1837–1872.
  55. Feature transfer learning for face recognition with under-represented data. In CVPR, 5704–5713.
  56. Cutmix: Regularization strategy to train strong classifiers with localizable features. In ICCV, 6023–6032.
  57. Pure Noise to the Rescue of Insufficient Data: Improving Imbalanced Classification by Training on Random Noise Images. In ICML, volume 162, 25817–25833.
  58. mixup: Beyond empirical risk minimization. ICLR.
  59. Deep Long-Tailed Learning: A Survey. IEEE TPAMI, 1–20.
  60. Bag of Tricks for Long-Tailed Visual Recognition with Deep Convolutional Neural Networks. In AAAI, 3447–3455.
  61. Improving Calibration for Long-Tailed Recognition. In CVPR, 16489–16498.
  62. BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed Visual Recognition. In CVPR, 9719–9728.
  63. Learning Deep Features for Discriminative Localization. In CVPR, 2921–2929.
  64. Places: A 10 million image database for scene recognition. IEEE TPAMI, 40(6): 1452–1464.
Citations (13)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now
Github Logo Streamline Icon: https://streamlinehq.com

GitHub

  1. GitHub - Keke921/H2T: This is the example code for H2T (9 stars)